You ran the analysis, one participant scored 98 on a scale where everyone else sits between 20 and 60, and now the mean, the standard deviation and the regression slope all move when you remove that single row. Should it stay or go? Handling outliers in SPSS is a two-part job: detecting the unusual cases with a stated rule, and then deciding, case by case and with a reason, whether each one is an error, a legitimate extreme value or a sign that your model needs a different test.
This article is part of our guide How to Write a Master’s Thesis by the Mastermind PhD team and relates to our Statistical Analysis service. Sources last checked: November 2026.
This guide covers both parts: four detection methods with the SPSS menus, a decision table, the options short of deletion, and how to report what you did.
What is an outlier? An outlier is an observation that lies unusually far from the rest of the data, either on one variable (a univariate outlier) or in its combination of scores across several variables (a multivariate outlier). Whether it is a problem depends on its cause: data entry errors, cases from outside the target population and genuine but rare values call for different actions.
Why do outliers in SPSS matter for a thesis?
Means, standard deviations, Pearson correlations and least-squares regression all weight extreme values heavily. A single extreme case can create a significant correlation that does not exist in the other 199 participants, or hide one that does. Examiners know this, so a results chapter that never mentions data screening invites the question at the defense.
Aguinis, Gottfredson and Joo (2013) reviewed how outliers are handled in management research and found that many published studies define, identify and treat them inconsistently, often without saying what they did. Their main recommendation is transparency: state the definition, the detection rule and the action for every type of outlier.
How do you detect univariate outliers in SPSS with a boxplot?
The boxplot is the quickest screen for outliers in SPSS. Use Analyze > Descriptive Statistics > Explore, move the variable into the Dependent List, and in Statistics tick Outliers to get the five highest and lowest values with their case numbers.
IBM’s documentation for the EXAMINE command defines the markers precisely. The box length is the interquartile range (IQR). Values more than 1.5 IQRs but less than 3 IQRs from the end of the box are labeled outliers and drawn as circles; values more than 3 IQRs away are labeled extreme and drawn with an asterisk. Each marker carries the case number, so you can go straight to the row in Data View.
The boxplot rule does not assume normality, which makes it a good first look for skewed variables such as income, reaction times or counts.
📱 Ask a statistician about your data — Send a short description of your instrument and sample size on WhatsApp and get a suitable analysis plan. Every analysis comes with the raw SPSS file and a results report.
How do you use z-scores to find outliers in SPSS?
For roughly normal variables, standardize the scores: Analyze > Descriptive Statistics > Descriptives, select the variables and tick Save standardized values as variables. SPSS adds a new column (for example Zscore_burnout) to the data file.
Tabachnick and Fidell (2019) treat cases with |z| greater than 3.29 (p < .001, two-tailed) as potential univariate outliers. Sort the new column or use Data > Select Cases with the condition ABS(Zscore_burnout) > 3.29 to list them.
One weakness: the mean and standard deviation used to compute z are themselves pulled by the outliers, so a very extreme case can partly mask itself and its neighbors.
What is the MAD rule and why is it more robust?
Leys, Ley, Klein, Bernard and Licata (2013) argued against the mean plus or minus several standard deviations and recommended the median absolute deviation (MAD), which is not distorted by the outliers it is trying to find. The steps:
- Compute the median of the variable.
- Compute the absolute deviation of every score from that median, then the median of those deviations.
- Multiply that median by 1.4826 to obtain the MAD.
- Flag cases more than 2.5 MADs from the median (their suggested default; 3 is very conservative, 2 is poorly conservative).
In SPSS you can do this with Transform > Compute Variable and Explore to read the medians. Whichever threshold you choose, decide it before looking at the results and state it in the method chapter.
How do you detect multivariate outliers in SPSS?
A case can be ordinary on each variable and still unusual in its combination, for example a young participant with very long work experience. For regression or MANOVA, check Mahalanobis distance:
- Open Analyze > Regression > Linear, enter any variable (such as the case ID) as the dependent and your predictors as independents.
- In Save, tick Mahalanobis (and Cook’s distance and leverage values if you are running the actual regression).
- Compare each saved distance with the chi-square critical value at p < .001, with degrees of freedom equal to the number of predictors. For three predictors the critical value is 16.27.
Cook’s distance shows influence, how much the regression coefficients change when the case is removed. A case can be an outlier without much influence, and an influential case is the one that deserves the most attention.
Which detection method should you use?
| Method | Best for | SPSS route | Typical cutoff |
|---|---|---|---|
| Boxplot (Explore) | Any distribution, a first visual screen | Analyze > Descriptive Statistics > Explore | 1.5 IQR (outlier), 3 IQR (extreme) |
| z-scores | Roughly normal variables, larger samples | Descriptives > Save standardized values | abs(z) > 3.29 |
| MAD rule | Skewed data, small samples, several outliers | Explore + Compute Variable | 2.5 MADs from the median |
| Mahalanobis distance | Multivariate outliers before regression or MANOVA | Linear Regression > Save > Mahalanobis | chi-square at p < .001 |
| Cook’s distance | Influence on regression results | Linear Regression > Save > Cook’s | values that stand apart from the rest; check above 1 |
Use one primary rule for outliers in SPSS per analysis and report it. Running every method until something is flagged, or until nothing is, is a form of analytic flexibility examiners recognize.
What should you do with outliers once you find them?
Start with the cause, not the statistic: outliers in SPSS are flagged by a rule, but kept or removed for a reason.
| Cause | Example | Action |
|---|---|---|
| Data entry error | Age entered as 220 instead of 22 | Correct it from the original questionnaire; if impossible, set it to missing |
| Out-of-range code | A 7 on a 1 to 5 Likert item | Treat as an entry error; check the paper or online record |
| Case outside the population | A manager in a sample defined as teachers | Exclude, and report the inclusion criterion |
| Careless responding | Same answer to every item, completed in 90 seconds | Exclude using a rule set in advance, and report it |
| Genuine extreme value | A real income far above the rest | Keep it; consider a transformation or a robust test |
For genuine values, the options short of deletion are: transforming the variable (log or square root for positive skew), winsorizing (replacing the extreme values with the nearest acceptable value, which should be reported), using a nonparametric test such as Mann-Whitney U or Spearman’s correlation, or running the analysis with and without the cases as a sensitivity check. If the conclusions are the same both ways, say so; that reassures the reader more than any single cutoff.
How do you report outliers in SPSS in the results chapter?
Put the screening of outliers in SPSS in the data preparation section, before the descriptive statistics. A typical scenario (not a real study):
Univariate outliers were screened with boxplots and standardized scores (abs(z) > 3.29). Two cases had entry errors (ages of 220 and 330), which were corrected from the original forms. Mahalanobis distance identified one multivariate outlier (chi-square(3) = 19.42, p < .001); results with and without this case did not differ in direction or significance, so it was retained. The final sample was 212.
That paragraph answers the four questions an examiner will ask: which rule, how many cases, what action and whether the result depends on it. Keep the syntax that produced the flags, so the decision is reproducible.
If you would like a statistician to review your data screening, the Mastermind PhD team delivers the raw SPSS file with every analysis. Statistical analysis starts from EGP 10,500 on the live service page, and the final quote depends on the number of variables and tests.
📱 Ask on WhatsApp now — 5 years of experience across all disciplines and 14 academic services. Two-stage payment (80/20); revisions are free until supervisor approval, within 30% of the chapter.
Frequently asked questions
Should I always remove outliers in SPSS?
No. Remove or correct a case only when you can name the cause: an entry error, a case outside the population or invalid responding. Genuine extreme values normally stay, with a transformation, a robust test or a sensitivity analysis.
What z-score counts as an outlier?
A common criterion is an absolute z above 3.29, which corresponds to p < .001 two-tailed (Tabachnick and Fidell). Some fields use 3 or 2.5. Choose the cutoff before screening and report it.
Do I check outliers before or after testing normality?
Before. Outliers distort skewness, kurtosis and the Shapiro-Wilk test, so screen and resolve them first, then assess normality on the cleaned variable.
How many outliers are too many?
There is no fixed percentage. If many cases are flagged, the variable is probably skewed rather than contaminated, and a transformation or a nonparametric test is more appropriate than deleting cases.
Can I find outliers in SPSS for Likert scale data?
Yes, but on the total or mean scale score rather than single items, where values are bounded. On single items, look for out-of-range codes and straight-lining patterns instead.
Sources
- Aguinis, H., Gottfredson, R. K., & Joo, H. (2013). Best-practice recommendations for defining, identifying, and handling outliers. Organizational Research Methods, 16(2), 270-301. https://doi.org/10.1177/1094428112470848
- Leys, C., Ley, C., Klein, O., Bernard, P., & Licata, L. (2013). Detecting outliers: Do not use standard deviation around the mean, use absolute deviation around the median. Journal of Experimental Social Psychology, 49(4), 764-766. https://doi.org/10.1016/j.jesp.2013.03.013
- Tabachnick, B. G., & Fidell, L. S. (2019). Using multivariate statistics (7th ed.). Pearson.
- IBM. PLOT Subcommand (EXAMINE command), IBM SPSS Statistics 27 documentation. https://www.ibm.com/docs/en/SSLVMB_27.0.0/statistics_reference_project_ddita/spss/base/syn_examine_plot.html (accessed November 2026)
- Field, A. (2024). Discovering statistics using IBM SPSS Statistics (6th ed.). SAGE.
Sources last checked: November 2026. This article is general educational guidance and does not replace your university regulations or your supervisor’s instructions. How we prepare our content.